This is the authors’ version of a paper that was later published as: Worthington, Andrew and Higgs, Helen (2004) Art as an investment: Risk, return and portfolio diversification in major painting markets. Accounting and Finance 44(2):pp. 257-272. Copyright 2004 Blackwell Publishing

Art as an investment: Risk, return and portfolio diversification in major painting markets

Andrew C. Worthington, Helen Higgs• School of Economics and Finance, Queensland University of Technology, Brisbane QLD 4001, Australia.

Abstract

This paper examines risk, return and the prospects for portfolio diversification among major painting and financial markets over the period 1976-2001. The art markets examined are Contemporary Masters, French Impressionists, Modern European, 19th Century European, Old Masters, Surrealists, 20th Century English and Modern US paintings. The financial markets comprise US Treasury bills, corporate and government bonds and small and large company stocks. In common with the literature in this area, the study finds that the returns on paintings are much lower and the risks much higher than conventional investment markets. Moreover, while low correlations of returns suggest that opportunities for portfolio diversification in art works alone and in conjunction with equity markets exist, the construction of Markowitz mean-variance efficient portfolios indicates that no diversification gains are provided by art in financial asset portfolios. However, diversification benefits in portfolios comprised solely of art works are possible, with Contemporary Masters, 19th Century European, Old Masters and 20th Century English paintings dominating the efficient frontier during the period in question.

Key words: Art and collectibles; Risk and return; Markowitz efficient frontier; Portfolio diversification

JEL classification: C61; D81; G11.

1. Introduction

In March 1987 Vincent Van Gogh’s [1853-1890] Sunflowers sold at auction for $39.9

million (all dollar figures in USD), followed in November by the sale of Irises for $53.9

million. Additional record-breaking sales in art markets followed closely. In May 1989, Pablo

Picasso’s [1881-1973] Yo Picasso sold for $47.8 million, far exceeding the $5.8 million that it

last commanded in May 1981: his Noces de Pierette later sold for $60.0 million. In May

The authors would like to thank seminar participants at the Queensland University of Technology and Massey University, Bruce Grundy (Associate Editor) and two anonymous referees for helpful comments on an earlier version of this paper. The financial support of an Australian Technology Network (ATN) Research Grant is also gratefully acknowledged.

Corresponding author: Associate Professor Andrew Worthington, School of Economics and Finance, Queensland University of Technology, GPO Box 2434, Brisbane QLD 4001, Australia, email. a.worthington@qut.edu.au.

2

1990, Van Gogh’s Portrait of Docteur Gachet sold for $82.5 million and Pierre-Auguste

Renoir’s [1841-1919] At the Moulin de la Galette for $78.1 million, becoming the two most

expensive pictures ever sold at auction (Pesando and Shum, 1999). Indeed, even demand for

Modern and contemporary paintings in the 1980s was so strong that works by (often still-

living) artists such as Roy Lichtenstein, Jackson Pollock, Jasper Johns, Robert Rauschenberg,

Willem de Kooning and Andy Warhol were frequently attracting prices in excess of $10

million (Anonymous, 2000). As a result, the international market for auctioned art grew

dramatically during this time, from less than $150 million in 1970 to more than $1.8 billion in

1997 (Renneboog and Van Houtte, 2000).

Obviously, these examples and others indicate that at least some paintings have increased

significantly in value and thereby generated large rates of return for their owners. For

instance, Irises had last been bought in 1948 for $84,000 (some $0.5 million at 1989 prices),

such that the record 1987 sale provided a 12 percent annual real rate of return (De la Barre et

al., 1994). And despite the well publicised bear market in art during the period 1989-1992,

there has been a sustained revival in picture prices over much of the last decade, especially in

areas outside the sky-high prices of Impressionist, Modern and contemporary works a decade

earlier (Curry, 1998). For example, Old Master sale prices have been rising strongly, with

many paintings selling for more than their high estimates. A rediscovered El Greco [1541-

1614] of The Crucifixion recently sold for more than six times its estimate at ₤3.6 million and

a tiny flower painting by Ambrosius Bosschaert the Elder [1573-1621] realised five times

expectations at ₤1.92 million (Anonymous, 2000).

In response to the commonly held belief that the art market yields huge profits in

comparison to other more prosaic investment markets, a small but growing literature has

examined the financial characteristics of the market in paintings, and art markets in general

(paintings, sculpture, ceramics and prints, along with collectibles such as coins, stamps,

antiques and furniture). This invariably accompanies a revival of interest in art investment by

the business world (see Oleck and Dunkin, 1999; Peers and Jeffrey, 1999). Starting with the

seminal work of Baumol (1986) much of this has been concerned with measuring the rate of

return of paintings (see, for example, Frey and Pommerehne, 1989; Buelens and Ginsburgh,

1993; Goetzmann, 1993; Pesando, 1993; Chanel et al., 1994; Frey and Eichenberger, 1995a,

1995b; Guerzoni, 1995; Candela and Scorcu, 1997; Frey and Pommerehne, 1998; Pesando

and Shum, 1999), however in recent years there has been an emerging emphasis on other

analytical dimensions of art investment (Felton, 1998).

3

Importantly, if art is to be regarded as a valid (albeit imperfect) addition to traditional

investments such as stocks and bonds amongst others, there is the requirement of examining

the prospects for diversification in such portfolios (Flores et al., 1998). Moreover, it is also

desirous to examine the prospects for diversification in portfolios composed primarily of art

held by investors, collectors, dealers and museums, amongst others. If low correlations of

returns exist between various art markets, diversifying across these markets may allow

investors to reduce portfolio risk while holding expected return constant. Unfortunately, little

empirical evidence currently exists concerning correlation among the differing art markets

and the concomitant prospects for portfolio diversification, both among art works alone and in

combination with financial assets.

Of course, it goes without saying that art markets differ substantially from financial

markets, and this potentially limits the strict applicability of well-known financial techniques.

Art works are not very liquid assets, almost never divisible, transaction costs are high, and

there are lengthy delays between the decision to sell and actual sale. Investing in art typically

requires extensive knowledge of art and the art world, and a large amount of capital to acquire

the work of well-known artists. The market is highly segmented and dominated by a few large

auction houses, and only a small number of works are presented for sale throughout the year.

Risk is also pervasive, deriving from both the physical risks of fire and theft and the

possibility of reattribution to a different artist, and the cost of insurance as a result can be

prohibitive. And while auction prices represent, in part, a consensus opinion on the value of

art works, values in turn are determined by a complex and subjective set of beliefs based on

past, present and future prices, individual tastes and changing fashion.

Nevertheless, in recent years it has been widely accepted that most art markets have moved

closer to the ideals set by financial markets. Turnover, for example, has increased

dramatically among the auction houses and the larger proportion of transactions are pursued

in these as against traditional dealers. Information on alternative art investments is now more

accessible through the attention of the media, and the publishing and dissemination of

catalogues and price index series has increased the amount of information available to both

buyers and sellers. Likewise, art markets are increasingly globalised, and the widening of the

asset pool to include collectibles, furniture, jewellery and wine, amongst others, has seen

substantially greater participation in most art markets. Of course, caution must play its part

when applying the tools of financial analysis to these scenarios, but such cross-disciplinary

research can still offer valuable insights, even with qualification.

4

The purpose of this paper is to examine risk, return and portfolio diversification in major

painting markets. Such information will provide important information to be both holders of

exclusively art portfolios and investors desiring to include art within mixed asset portfolios

comprising stocks and bonds, amongst others. The paper itself is divided into four main areas.

Section 2 explains the data and empirical methodology employed in the present analysis.

Section 3 discusses risk and return in art markets while Section 4 presents an analysis of

portfolio diversification. The paper ends with some brief concluding remarks in the final

section.

2. Data and methodology

The data employed in the study is composed of indices for nine major categories of

paintings and five financial markets. All art index data is obtained from UK-based Art Market

Research (AMR) (2003) and encompass the period January 1976 to December 2001. AMR art

indexes are widely used by a variety of leading institutions concerned with prices in the arts,

including Christie’s, Sotheby’s, the British Inland Revenue Service and the New York Federal

Reserve, along with the Financial Times, Wall Street Journal, The Economist, Business Week,

The Art Newspaper and Handelsblatt (AMR, 2003) [see, for example, Anonymous (2000)]. In

brief, the indices are calculated by collecting all worldwide sales in each month by artist,

converting these values to US dollars, trimming by ten percent to eliminate extreme values

and calculating the ratio to the January 1976 base period. For schools, movements and

periods, the individual artists are combined together to form an equal-weighted portfolio.

The nine major art indexes are as follows: (i) Contemporary Masters (CM), covering 5,106

sales of current masters including Basquiat, Clemente and Polke; (ii) 20th Century English

(TE) encompassing 10,603 sales by artists such as Dawson, Flint, Moore and Munnings; (iii)

19th Century European (NE) with 50,510 sales by artists including Maris, Troyon, Constable

and Corot; (iv) French Impressionist (FI) with sales of 6,242 works by painters including

Degas, Monet and Renoir; (v) Modern European (ME) with 17,538 sales by artists like

Bonnard, Picasso and Utrillo; (vi) Modern US Paintings (US) with 10,607 sales of works by

painters such as Kooning, Rivers and Warhol; (vii) Old Masters (OM) with 6,412 sales by

artists including Gainsborough, Reynolds and Storck; (viii) Surrealists (SR) with 10,395 sales

by artists including Dali, Magritte and Picabia; and a general painting (ART) index with

94,514 sales by a selection of one hundred artists across all schools and periods including

artists such as Boyd, Foster and Rivera. All data is monthly and specified in US dollar terms.

5

The indexes selected are consistent with studies in the area of art investment returns and

risk and represent some of the most closely followed painting sub-sectors. It is also important

to note that the aggregation of these art indices across schools, movements and periods

produces a portfolio diversification effect when compared to artists in much the same manner

as the aggregation of companies in industries and markets. This effect varies from index to

index. For example, French Impressionists (FI) are drawn from a small number of artists from

a single movement and a relatively narrow period (mostly Degas, Monet, Renoir and Pissarro

from the 1860s until Post-Impressionism), whereas Old Masters (OM) comprises a larger

number of artists drawn from several movements and periods (Renaissance, Mannerist and

Baroque, especially Italian, Dutch and Flemish artists, largely pre-1700) [see AMR (2003) for

details].

The five financial market indices represent major asset classes and are defined as: large

company stocks (LCS), small company stocks (SCS), long-term corporate bonds (LCB), long-

term government bonds (LGB) and treasury bills (TBL). All Ibbotson Associates (2002)

indices are constructed using US equity and bond markets and this maintains consistency with

the art data as defined in USD terms and with the position of the US as the world’s leading art

market [some 44 percent of the global art market followed by the UK with 29 percent

(Renneboog and Van Houtte (2002)]. The series employed are total monthly returns, and

where applicable, include capital appreciation and income. For both the art and financial

indices the monthly returns are calculated such that the monthly return in market i is

represented by the continuously compounded return or log return of the price index at time t

such that ( ) 100log 1 ×=Δ −ititit ppp where Δpit denotes the rate of change of pit.

The analysis of art risk, return and portfolio diversification in the paper is made as

follows. To start with, two different data sets are examined. The first set of data is returns for

the eight individual schools or periods of paintings and is examined in the context of an

exclusively art portfolio. That is, CM, FI, ME, NE, OM, SR, TE and US. The second set of

data relates to a broad art asset included in a multi-asset class portfolio with other financial

assets: namely, ART, LCS, SCS, LCB, LGB and TBL. Two stages of analysis are followed in

each case. First, the central tendency, dispersion and shape of all series are examined, along

with their time series properties. Second, following Markowitz’s (1952) well-known portfolio

theory, combinations of these assets with different risk-return characteristics are constructed.

Within this set of all possible combinations, the set of portfolio strategies with the least

variance for a given mean return produces the mean-variance frontier (also known as the

6

mean-variance set). The mean-variance frontier is then further identified as an efficient

frontier (efficient set) representing portfolios where portfolio return is maximized for a given

level of portfolio risk. Portfolios are constructed employing both optimizing (mean-variance

efficient investment) and naïve (single market and equally spread investment) strategies. One

constraint placed on asset allocation within the portfolios is that negative (or short) positions

are not allowable in order to correctly reflect the realities of investment in art markets. For

simplification, cash is assumed to be zero

3. Risk and return

Table 1 presents descriptive statistics for the eight individual art markets. Arithmetic and

geometric means, medians, maximums, minimums, standard deviations, coefficients of

variation, risk-adjusted returns, skewness, kurtosis, Jacque-Bera statistics and p-values and

Augmented Dickey-Fuller (ADF) unit root tests are reported. Within the period examined, the

arithmetic mean annual returns for the art markets range from 1.90 percent for Surrealists (SR)

to 4.22 percent for Contemporary Masters (CM). The geometric mean annual returns are all

lower than the arithmetic means for each market, suggestive of the high volatility in returns

over the period examined. The highest geometric mean returns are in Contemporary Masters

(CM) and 20th Century English (TE) with 3.73 and 2.85 percent, respectively. The lowest

geometric mean annual returns are in Modern European (ME) (1.49 percent) and Surrealists

(SR) (1.25 percent). Risk, as measured by standard deviation of art returns, ranges from 7.24

percent for 20th Century English (TE) up to 13.86 percent for French Impressionists (FI). Of

the eight art markets 20th Century English (TE) and 19th Century European (NE) are the least

volatile, while French Impressionist (FI) and Modern US Paintings (US) are the most volatile.

<TABLE 1 HERE>

In terms of the relationships between risk and return, two measures are calculated and

presented in Table 1. First, the value of the coefficient of variation (standard deviation divided

by the mean return) measures the degree of risk in relation to return. The lowest coefficients

of variation are 20th Century English (TE) (2.34) and Contemporary Masters (CM) (2.48). The

Surrealists (SR) has the highest coefficient of variation of 5.96. Second, risk-adjusted returns

are calculated using return divided by standard deviation in order to measure return in relation

to risk. The results illustrate the dominance in risk-adjusted returns of 20th Century English

7

(TE) (0.43) and Contemporary Masters (CM) (0.40) over Surrealists (SR) (0.17) and Modern

European (ME) (0.19) paintings.

To evaluate the shapes of the distributions, skewness, kurtosis and Jarque-Bera statistics

are also presented in Table 1. Since the sampling distribution of skewness is normal with

mean 0 and standard deviation of T6 , where T is the sample size, then Contemporary

Masters (CM) is significantly positively skewed at the .05 level (that is, with unusually high

returns) and French Impressionists (FI), Modern European (ME), Surrealists (SR) and 20th

Century English (TE) are significantly negatively skewed (with unusually low returns). For

kurtosis the sampling distribution is also normal with mean 0 and standard deviation of

T24 . Accordingly, Contemporary Masters (CM), French Impressionists (FI), 19th Century

European (NE), and Surrealists (SR) markets with kurtosis exceeding three can be represented

by a leptokurtic distribution, whereas Modern European (ME), Old Masters (OM), 20th

Century English (TE) and Modern US Paintings (US) with kurtosis of less than three have a

platykurtic distribution.

The Jarque-Bera statistic and p-values in Table 1 are used to test the null hypothesis that

the distribution for the art market returns is normally distributed. All p-values are greater than

the 0.01 level of significance indicating that art returns can be approximated by a normal

distribution with the exception of French Impressionists (FI). Finally, ADF unit root tests are

calculated for the eight series of painting returns. In all instances, the null hypothesis of

nonstationarity is tested. The analyses of the art market returns at levels indicate all painting

markets are stationary at the 0.01 level of significance.

The correlation matrix for the eight art markets is included in the lower portion of Table 1.

All pairwise correlations are positive and range from 0.41 to 0.86. French Impressionists (FI)

shows a high correlation with 19th Century European (NE) (0.86), while here is also a high

positive correlation of 0.83 between Contemporary Masters (CM) and Surrealists (SR). The

lowest positive correlations are 0.41 between Modern European (ME) and Old Masters (OM)

and 0.41 between 20th Century English (TE) and Modern US Paintings (US). The finding that

the art markets are not perfectly positively correlated is suggestive of the potential benefits of

portfolio diversification in exclusively art portfolios.

Table 2 depicts the summary of descriptive statistics for the returns of large company

stocks (LCS), small company stocks (SCS), long-term corporate bonds (LCB), long-term

government bonds (LGB), treasury bills (TBL) and the art market (ART). The arithmetic

average annual returns for the six markets range from 3.03 percent for the art market (ART) to

8

17.86 percent for small company stocks (SCS). The lowest mean return next to art is treasury

bills (TBL) (6.52 percent). Once again, the geometric means are all lower than the arithmetic

means for each market. The highest geometric average annual returns are in small company

stocks (SCS) and large company stocks (LCS) averaging 16.82 and 13.19 percent respectively.

The lowest geometric mean annual returns are in art markets (ART) and treasury bills (TBL)

with 2.54 and 6.49 percent, respectively. The standard deviations for the art, equity, bond and

bills range from 2.62 to 15.29. Small company stocks (SCS) have the highest return and risk,

while treasury bills (TBL) have the lowest return and risk. Of the six markets, small company

stocks (SCS) and large company stocks (LCS) are the most volatile, while treasury bills (TBL)

is the least volatile.

<TABLE 2 HERE>

The value of the coefficient of variation (standard deviation divided by the mean return)

measures the degree of risk in relation to the mean return. The lowest coefficients of variation

are treasury bills (TBL) (0.40) with very low risk and return and highest are in small company

stocks (SCS) (0.86) with high risk and return. The art market (ART) has the highest coefficient

of variation (3.34) with a very low arithmetic mean return relative to risk (as measured by

standard deviation). There is a clear dominance in risk-adjusted returns of treasury bills (TBL)

(2.49) and small company stocks (SCS) (1.17) over art markets (ART) (0.30) and long-term

government bonds (LGB) (0.83).

In terms of skewness, all the asset classes are significantly skewed. For kurtosis, small

company stocks (SCS), long-term corporate bonds (LCB) and treasury bills (TBL) can be

represented by a leptokurtic distribution whereas art (ART), large company stocks (LCS) and

long term government bonds (LGB) with kurtosis of less than three can be represented by a

platykurtic distribution. The Jarque-Bera statistic and corresponding p-values in Table 2 fail

to reject the null hypothesis that the distributions are normally distributed. The ADF unit root

tests are presented for the six markets in the returns series. The analyses of the art, equity and

bond market returns at levels indicate that returns in all markets are stationary with the

exception of art (ART) and treasury bills (TBL), which are stationary in differences.

Finally, Table 2 also presents the correlation matrix for the six markets. The pairwise

correlations range from -0.3058 to 0.9629 and are generally positive. The exceptions are that

art markets (ART) are negatively correlated with small company stocks (SCS) and long-term

corporate bonds (LCB) while long-term government bonds (LGB) and small company stocks

(SCS) are also negatively correlated. The low correlations between art markets and most of

9

the financial markets are once again suggestive of the potential gains for portfolio

diversification involving art investment.

4. Portfolio diversification

In the second part of the analysis, Markowitz portfolio theory is used to construct the

efficient frontier for the exclusively art portfolio and the mixed asset portfolio where art is

included alongside short and long term government debt and corporate debt and equity.

Mean-variance portfolio optimisation is made using the Microsoft Excel-based program M-V

Optimiser (Wagner Associates, 2003). Figure 1 depicts the efficient frontier derived from the

various combinations of the eight art markets. One hundred and four different portfolios are

included in the efficient set. Also included with the mean-variance frontier are several naïve

portfolios that are either art investments made in a single market (CM, FI, ME, NE, OM, SR,

TE and US) or an equally weighted portfolio across all markets (EQW).

<FIGURE 1 HERE>

The returns (risks) for the efficient frontier range from 2.71 percent (6.44 percent) at the

minimum variance point to 4.22 percent (10.47) at its uppermost. All other things being equal,

naïve strategies, where investment is made solely in one art market or equally in all markets,

are dominated by the efficient set. Only Contemporary Masters (CM), with the highest risk

and return, lies on the efficient frontier. However, due to the relatively narrow range of

returns in the feasible set of mean-variance portfolios, even the dominance of the efficient

frontier over the naïve strategies is small, such that the return of the equal weighting strategy

is only 0.015 percent less than the highest return in the efficient set.

As often as not, it is expected that individual assets that are plotted farthest from the

efficient frontier are excluded from the set of efficient portfolios and this is indeed the case

with the naïve strategies of investing in French Impressionists (FI) and Modern US Paintings

(US) alone. In fact, the efficient frontier is mostly comprised of just two or three of the eight

art assets included in the calculations. For example, when return is equal to 2.71 percent

(Point A) the only assets included (with their portfolio weight) are 19th Century European

(NE) (21.12 percent), Old Masters (OM) (27.45 percent) and 20th Century English (TE) (51.43

percent). At Point B (3.16 percent return) the frontier point is composed of Contemporary

Masters (CM) (19.27 percent), Old Masters (OM) (20.38 percent) and 20th Century English

(TE) (60.35 percent). At Point C Contemporary Masters (CM) (52.74 percent) and 20th

10

Century English (TE) (47.26 percent) form the efficient portfolio, changing to 68.94 and

31.06 percent at Point D. Finally, at Point E the efficient portfolio is composed solely of

Contemporary Masters (CM).

While the parameters used in constructing the efficient frontier for art investment are

historical, there are a number of points to make regarding portfolio diversification. To start

with, over the period 1976 to 2001 a relatively small number of separate art markets dominate

the efficient set, mainly 19th Century European (NE), Old Masters (OM), 20th Century English

(TE) and Contemporary Masters (CM). Moreover, Contemporary Masters (CM) with its

relatively high returns over the period generally dominates portfolios with high risk-return

characteristics, while 20th Century English (TE) dominates portfolios with low risk-return

characteristics. Accordingly, taking as given an art investor’s subjective risk preferences, less

risk averse investors would tend to favour Contemporary Masters (CM) in their diversified

portfolios, while more risk averse investors would tend to weight heavily in favour of 20th

Century English (TE) works. Of the remaining painting markets, French Impressionists (FI),

Surrealists (SR), Modern US Paintings (US) and Modern European (ME) painting are

generally not included in the efficient set through their high risk-low return characteristics

over the period in question. In general, it would appear that most of the gains from

diversification achievable in art can be made with a small number of individual painting

markets, though of course the performance of individual artists and schools within these

markets could vary markedly from the market as a whole.

To construct the efficient frontier in Table 2, a general art asset class (ART) is included

alongside short and long term government debt (TBL and LGB) and corporate debt (LCB) and

equity (SCS and LCS). Six portfolios employing a naïve strategy are again plotted, being

investment in a single art, bond or equity market or in an equally weighted portfolio (EQW),

along with one hundred and five portfolios that form the efficient frontier. The returns (risks)

for the efficient frontier of the art and equity markets vary from 6.87 percent (2.54) to 17.86

percent (15.29). Importantly, and unlike the experience with the exclusively art portfolio,

most of the assets in the mixed-asset portfolio are included in some way in the efficient set.

For instance, Treasury bills (TBL) average 34.54 percent weighting across all efficient

portfolios, small company stocks (SCS) 44.69 percent, large company stocks (LCS) 10.43

percent and long-term corporate (LCB) and government (LGB) bonds 5.04 and 5.29 percent,

respectively. Nonetheless, at points A and B (low risk-return) the portfolio is dominated by

Treasury bills (TBL) and at C, D and E (high risk-return) by small company stocks (SCS). In

11

most cases, the mean-variance efficient portfolios out perform the naïve strategies, and in

some the gains are quite substantial.

<FIGURE 2 HERE>

However, in none of the risk-return optimal portfolios that define the efficient frontier is

the art market included. It is clear that even though art markets have very low and even

negative correlations with financial market assets, their risk-return characteristics are so

inferior to equity and debt markets that they are never included in the efficient set. This would

suggest, for the most part, that the diversification benefits of art in a multi-financial asset

portfolio are close to zero. Renneboog and Van Houtte (2002, p. 349) likewise concluded that

“the Markowitz efficient frontier of an investment opportunity set consisting of both equity

and art investment does not shift upwards…suggesting that the diversification potential of art

in an equity setting is limited”. However, it is the case that as the number of assets increases

the risk of the portfolio collapses to the individual covariances, such that the creation of a

portfolio with much finer detail than the broad asset classes used here should illustrate at least

some diversification benefits.

5. Summary and conclusions

This paper investigates risk, return and portfolio diversification in major painting markets

during the period 1976 to 2001. The art markets examined are Contemporary Masters, French

Impressionists, Modern European, 19th Century European, Old Masters, Surrealists, 20th

Century English and Modern US paintings. Comparison is also made between art and

financial assets, including US government debt and corporate debt and equity. In common

with most other work in this area, the results indicate that the returns to art investment is less,

and the risks much higher, than in more conventional investment markets. However, there is

also much variation across the different art markets, with Contemporary Masters, 19th Century

European, 20th Century English and Old Masters offering superior risk-adjusted returns (at

least among painting markets) during the period question.

At first impression, the low correlations of returns among art works and between art works

and financial assets are suggestive of the benefits of portfolio diversification. Certainly, gains

to portfolio diversification exist in the first respect, and this offers potential guidance for

portfolios composed primarily of art held by investors, collectors, dealers and museums,

amongst others. However, it is also the case that the risk-return attributes of art are so inferior

12

to financial assets such as stocks and bonds during the period in question that inclusion of

these assets for diversification purposes in financial asset portfolios cannot be supported, at

least in a portfolio composed of the broad asset classes as used here. Renneboog and Van

Houtte (2000, 2002) also concluded that Markowitz efficient frontiers showed limited

diversification potential for art, though just in the context of equity investment. Of course, the

art returns as calculated do not reflect the fact that a substantial component of the return from

art investment can be derived not from financial returns, rather its intrinsic aesthetic qualities.

Equally, they also do not include the many and sizeable transaction and holding costs

associated with art portfolios, the absence of which may serve to inflate financial returns.

There are many interesting opportunities to expand upon this work by applying some of the

well-used tools of financial analysis to art markets. One possibility is to follow the work of

Ginsburgh and Jeanfils (1995), Chanel (1995) and Czujack et al. (1996) and more closely

examine the short and long run relationships between art and financial markets. A few studies

have already investigated the wealth effect link between stock and art markets and concluded

that booms in stock markets create booms in art markets (see, for instance, Goetzmann, 1993;

Chanel, 1995), However, the exact strength and persistence of this causal relationship is less

well known. Another avenue for research would be to examine the opportunities for arbitrage

that exists between different geographical markets (say, New York, London and Paris). This

would permit greater empirical certainty on the global efficiency of art markets. Asset-pricing

models could also be applied to artworks in order to gain a greater awareness of price

formation in these markets. A hedonic pricing equation as followed in Renneboog and Van

Houtte (2002) is one option; another is Zanola and Locatelli Biey’s (1998) short-run capital

asset pricing model approach. Finally, there may be potential to examine art markets along the

lines of the momentum-investing literature. Investor over and under reaction may go far in

explaining the sustained bull market in art up to 1989 and the bear market in the following

three years.

References

Anonymous, 2000, New century, old masters, The Economist, 356, 84. Art Market Research, 2003, <http://www.artmarketresearch.com/>, Accessed March 2003. Baumol, W.J., 1986, Unnatural value: Or art investment as a floating crap game, American Economic Review,

76, 10–14. Buelens, N. and V. Ginsburgh, 1993, Revisiting Baumol’s ‘Art as a floating crap game’, European Economic

Review, 37, 1351–1371. Candela, G. and A.E. Scorcu, 1997, A price index for art market auctions, Journal of Cultural Economics, 21,

175–196. Chanel, O., 1995, Is art market behaviour predictable?, European Economic Review, 39, 519–527. Chanel, O., L.A. Gerard-Varet, and V. Ginsburgh, 1994, Prices and returns on paintings: An exercise on how to

13

price the priceless, Geneva Papers on Risk and Insurance Theory, 19, 7–21. Curry, J., 1998, Art as an alternative investment, Trust and Estates, 137, 25–26. Czujack, C., R. Flores, and V. Ginsburgh, 1996, On long-run price comovements between paintings and prints,

in: V.A. Ginsburgh and P.M. Menger, eds., Economics of the arts: selected essays (Elsevier North-Holland, Amsterdam) 85–112.

De la Barre, M., S. Docclo, and V. Ginsburgh, 1994, Returns of impressionist, modern and contemporary European paintings, 1962-1991, Annales d’Economie et de Statistique, 35, 143–181.

Felton, M.V., 1998, Review of: Economics of the arts: Selected essays, Journal of Economic Literature, 36, 286–287.

Flores, R.G., V. Ginsburgh, and P. Jeanfils, 1998, Long and short term portfolio choices of paintings, Journal of Cultural Economics, 23, 193–201.

Frey, B.S. and R. Eichenberger, 1995a, On the return of art investment return analyses, Journal of Cultural Economics, 19, 207–220.

Frey, B.S. and R. Einchenberger, 1995b, On the rate of return in the art market: Survey and evaluation, European Economic Review, 39, 528–537.

Frey, B.S. and W.W. Pommerehne, 1998, Art investment: An empirical inquiry, Southern Economic Journal, 56, 396–409.

Ginsburg, V. and P. Jeanfils, 1995, Long term comovements in international markets for paintings, European Economic Review, 39, 538–548.

Goetzmann, W.N., 1993, Accounting for taste: Art and the finance markets over three centuries, American Economic Review, 83, 1370–1376.

Guerzoni, G., 1995, Reflections on historical series of art prices: Reitlinger’s data revisited, Journal of Cultural Economics, 19, 251–260.

Ibbotson Associates, 2002, Stocks, bonds, bills and inflation, 2001 yearbook (Ibbotson Associates: Chicago). Markowitz, H.M., 1952, Portfolio selection, Journal of Finance, 2, 184–199. Oleck, J. and A. Dunkin, 1999, The art of collecting art, Business Week, May 17. Peers, A. and Jeffrey N.A., 1999, Art and money, Wall Street Journal, Nov 12. Pesando, J.E. and P.M. Shum, 1999, The returns to Picasso’s prints and to traditional financial assets, 1977 to

1996, Journal of Cultural Economics, 23, 183–192. Pesando, J.E., 1993, Arts as an investment: The market for modern prints, American Economic Review, 83,

1075–1089. Renneboog, L. and T. Van Houtte, 2000, From realism to surrealism: Investing in Belgian art, Cahiers

Economiques de Bruxelles, 165, 69–106. Renneboog, L. and T. Van Houtte, 2002, The monetary appreciation of paintings: From realism to Magritte,

Cambridge Journal of Economics, 26, 331–357. Wagner Associates, 2003, <http:// www.wagner.com>, Accessed March 2003. Zanola, R. and M. Locatelli Biey, 1998, Speculative investment in painting: A short run capital asset price

index, Department of Economics Working Paper, University of York.

Table 1 Selected descriptive statistics for returns in eight painting markets, 1976-2001

CM FI ME NE OM SR TE US Arithmetic mean 0.0422 0.0331 0.0212 0.0223 0.0234 0.0190 0.0310 0.0284Geometric mean 0.0372 0.0229 0.0149 0.0197 0.0202 0.0125 0.0285 0.0203Median 0.0398 0.0606 0.0278 0.0118 0.0311 0.0275 0.0486 0.0155Maximum 0.2971 0.3420 0.2163 0.1703 0.1822 0.2268 0.1230 0.2647Minimum -0.1526 -0.4051 -0.2372 -0.1610 -0.1449 -0.2938 -0.1106 -0.2741Standard deviation 0.1047 0.1386 0.1126 0.0734 0.0826 0.1131 0.0724 0.1294Skewness 0.4868 -0.8303 -0.4252 -0.0376 0.0859 -0.6550 -0.5302 -0.0762Kurtosis 3.5073 5.5636 2.5876 3.2626 2.2067 3.8093 2.0266 2.7382Coefficient of variation 2.4790 4.1926 5.3053 3.2972 3.5256 5.9603 2.3357 4.5656Risk-adjusted return 0.4034 0.2385 0.1885 0.3033 0.2836 0.1678 0.4281 0.2190Jarque-Bera statistic 1.3057 10.1067 0.9677 0.0809 0.7138 2.5685 2.2448 0.0994

Cen

tral t

ende

ncy,

dis

pers

ion

& sh

ape

Jarque-Bera p-value 0.5206 0.0064 0.6164 0.9604 0.6999 0.2769 0.3255 0.9515ADF (Level) -4.5183 -4.5059 -4.4139 -4.0393 -4.7825 -4.9220 -4.1594 -4.4502Critical value .01 level -3.9888 -3.9888 -3.9888 -3.9888 -3.9888 -3.9888 -3.9888 -3.9888Critical value .05 level -3.4248 -3.4248 -3.4248 -3.4248 -3.4248 -3.4248 -3.4248 -3.4248

AD

F te

sts

Critical value .10 level -3.1355 -3.1355 -3.1355 -3.1355 -3.1355 -3.1355 -3.1355 -3.1355CM 1.0000 – – – – – – –

FI 0.7437 1.0000 – – – – – –ME 0.6926 0.7061 1.0000 – – – – –NE 0.7605 0.8605 0.7470 1.0000 – – – –

OM 0.5060 0.5451 0.4057 0.7330 1.0000 – – –SR 0.8270 0.6602 0.7632 0.6931 0.5113 1.0000 – –TE 0.5667 0.6392 0.7250 0.6520 0.4331 0.6143 1.0000 –

Cor

rela

tion

US 0.6567 0.5447 0.4456 0.5440 0.4387 0.4169 0.4135 1.0000Notes: Means, median, maximum and minimum are in annualised terms. CM – Contemporary Masters, FI – French Impressionists, ME – Modern European, NE – 19th Century European, OM – Old Masters, SR – Surrealists, TE – 20th Century English, US – Modern US Paintings. Critical values at the .05 level for skewness and kurtosis are 0.2718 and 0.5435, respectively. Risk adjusted return is the ratio of return to standard deviation; coefficient of variation is ratio of standard deviation to return. For Augmented Dickey-Fuller (ADF) tests hypotheses are H0: unit root, H1: no unit root (stationary). The lag orders in the ADF equations are determined by the significance of the coefficient for the lagged terms. Intercepts and trends are included in the levels series. Correlation is Pearson (product moment) correlation.

15

Table 2 Selected descriptive statistics for annualised returns in painting and financial markets, 1976-2001

ART LCS SCS LCB LGB TBL Arithmetic mean 0.0303 0.1395 0.1786 0.0963 0.0975 0.0652 Geometric mean 0.0254 0.1319 0.1682 0.0915 0.0916 0.0649 Median 0.0305 0.1698 0.2234 0.1040 0.0886 0.0560 Maximum 0.2281 0.3234 0.3928 0.3672 0.3487 0.1381 Minimum -0.2115 -0.1076 -0.2205 -0.0761 -0.0923 0.0286 Standard deviation 0.1012 0.1321 0.1529 0.1051 0.1168 0.0262 Skewness -0.3145 -0.3538 -0.7648 0.5256 0.3250 0.9758 Kurtosis 2.8978 1.9600 3.0898 3.1103 2.2580 3.5077 Coefficient of variation 3.3417 0.9463 0.8562 1.0912 1.1971 0.4011 Risk-adjusted return 0.2993 1.0567 1.1680 0.9165 0.8354 2.4929 Jarque-Bera statistic 0.4399 1.7142 2.5432 1.2101 1.0542 4.4056

Cen

tral t

ende

ncy,

dis

pers

ion

& sh

ape

Jarque-Bera p-value 0.8026 0.4244 0.2804 0.5460 0.5903 0.1105 ADF (Level) -3.9616 -4.9445 -5.9095 -4.8880 -5.1064 -2.8936 Critical value .01 level -3.9888 -3.9888 -3.9888 -3.9888 -3.9888 -3.9888 Critical value .05 level -3.4248 -3.4248 -3.4248 -3.4248 -3.4248 -3.4248 Critical value .10 level -3.1355 -3.1355 -3.1355 -3.1355 -3.1355 -3.1355 ADF (Difference) -6.0253 – – – – -5.7827 Critical value .01 level -3.4520 – – – – -3.4520 Critical value .05 level -2.8710 – – – – -2.8710

AD

F te

sts

Critical value .10 level -2.5719 – – – – -2.5719 ART 1.0000 – – – – – LCS 0.1618 1.0000 – – – – SCS -0.3058 0.3800 1.0000 – – – LCB -0.0724 0.3234 0.0175 1.0000 – – LGB 0.0220 0.3297 -0.0247 0.9629 1.0000 – C

orre

latio

n

TBL 0.3009 0.0871 0.0808 0.0261 0.0485 1.0000 Notes: Means, median, maximum and minimum are in annualised terms. ART – art market, LCS – large company stocks, SCS – small company stocks, LCB – long-term corporate bonds, LGB – long-term government bonds, TBL – Treasury bills. Critical values at the .05 level for skewness and kurtosis are 0.1387 and 0.2773, respectively. Risk adjusted return is the ratio of return to standard deviation; coefficient of variation is ratio of standard deviation to return. For Augmented Dickey-Fuller (ADF) tests hypotheses are H0: unit root, H1: no unit root (stationary). The lag orders in the ADF equations are determined by the significance of the coefficient for the lagged terms. Intercepts and trends are included in the levels series, intercepts only in the first-differenced series. Correlation is Pearson (product moment) correlation.

16

0.015

0.020

0.025

0.030

0.035

0.040

0.045

0.050

0.050 0.060 0.070 0.080 0.090 0.100 0.110 0.120 0.130 0.140 0.150

Risk

Ret

urn

CM

TE

EQW

FI

US

MEOMNE

SR

A

B

CD

E

Fig 1. Efficient frontier for art investments

0.000

0.020

0.040

0.060

0.080

0.100

0.120

0.140

0.160

0.180

0.200

0.020 0.040 0.060 0.080 0.100 0.120 0.140 0.160 0.180

Risk

Ret

urn

SCS

EQ

ART

LCB LGB

LCS

TBLA

B

C

D

E

Fig 2. Efficient frontier for art and financial investments